Compress PDF is a browser-only tool that reduces the size of a JPG-heavy PDF by re-encoding eligible embedded JPG images locally, without uploading your file. You open one PDF on your device, the tool walks through the document's image objects, and it only rewrites the JPG streams it can safely replace with smaller versions at a balanced quality setting. Everything else, including text, links, forms, annotations, and page structure, stays in place. Because the work happens inside your current browser tab, there is no upload step, no signup, and no server holding a copy of your document. That makes the tool a practical choice when you need to shrink a scanned or photo-heavy PDF for email, a chat attachment, or a form upload that has a size cap, and when you would rather not hand the file to a third-party service. The trade-off is honesty: if no embedded image can be made smaller without losing visual quality, the tool does not generate a download. You see the input and output sizes it measured, and you decide whether the saved bytes are worth keeping the result.
What Compress PDF actually does
Compress PDF is a deliberately narrow, browser-only optimizer. It does not try to be a general-purpose PDF compressor with stream-level filters, archival conformance, or print color management. Instead, it focuses on the single most common reason a PDF gets too large: image-heavy pages where the embedded JPG photos were saved at a high resolution or a high JPEG quality. A scanned contract, a brochure exported from a design tool, or a product catalogue saved for email are exactly the cases this tool was built for.
When you load a PDF, the tool walks through the document's image objects using pdf-lib. It allows only one image type to be replaced: unmasked DCTDecode JPG images using DeviceRGB or DeviceGray color data, with no special decode parameters. Anything else, including JPX (JPEG 2000), CMYK, Indexed color, masked images, ImageMask, and any image with decode parameters, is skipped and left exactly as it was. The tool is honest about this: it does not guess, and it does not silently rewrite something it is not confident about.
For each eligible image, the tool decodes the JPG inside a browser image element, draws it onto a canvas, and re-exports the canvas as a fresh JPG using the WHATWG canvas toBlob interface. The longest side of the image is capped at 1400 pixels, and the export is saved at JPEG quality 0.70. Those are the only two knobs, and they are balanced for everyday use: enough resolution to look fine on screen and in standard printing, and a quality setting that keeps file size honest without producing visible artifacts in normal viewing. A 1400-pixel-wide image is well above what most laptop screens and printers render anyway, so the resize is rarely visible to the reader.
After each re-export, the tool compares the new JPG byte stream against the original. If the candidate is not actually smaller, the original image is kept untouched. This check happens for every eligible image, so the output file is built out of only the genuinely smaller replacements. If no replacement is smaller anywhere in the document, the tool returns an honest error message and does not offer a download. It would rather show you nothing than hand you a renamed copy of the same file.
How to compress a PDF easily with Compress PDF
- Open Compress PDF in your browser and pick one local PDF that is no larger than 25 MB.
- Select Compress PDF and wait while the tool inspects the embedded image objects, decodes the eligible JPGs, and re-encodes them locally at 1400 px and JPEG quality 0.70.
- Read the reported input size and output size. If the output is smaller and the page count has been verified, a Download button appears.
- Click Download to save the verified PDF. Keep the original file while you inspect the result, especially before sharing it.
What stays the same in your PDF
The tool only swaps the byte payload of eligible JPG images, and it does so at the same indirect reference inside the PDF structure. That surgical approach means everything else about the document is preserved as it was:
- Page count and page order
- Text, fonts, and selectable copy
- Internal and external links
- Vector objects, drawings, and shapes
- Interactive form fields and widgets
- Annotations, comments, and metadata
- Any image that is not an unmasked DCTDecode JPG in DeviceRGB or DeviceGray
This matters because the output is still the same PDF; it just has a smaller body for the parts that were safe to rewrite. Pages are not rasterized, so text does not become blurry, and links do not break. Before the download button appears, the tool reopens the output with pdf-lib and PDF.js and checks that the page count matches the source. If the result is malformed at all, the tool does not present it as success; the download is suppressed.
When the tool shrinks a file and when it cannot
The compression you get depends on what is actually inside your PDF. The tool does not pad results or invent savings, so the outcome is directly tied to your source images and the PDF overhead around them.
| Document type | Likely result |
|---|---|
| Scanned pages saved as JPG in RGB or grayscale | Usually shrinks noticeably |
| Photo-heavy catalogue or brochure | Often shrinks a lot |
| Mixed PDF with text and a few JPG logos | Modest savings, depends on the image share of the file |
| Already-optimized PDF with small JPGs | May stay the same size; the tool keeps the original whenever a re-encode is not smaller |
| Text-only PDF or a PDF with no JPG images | Usually cannot shrink |
| PDF using only JPX, CMYK, Indexed, masked, or decode-parameter images | No change; these image types are skipped |
That last row matters. A PDF created from a scanner sometimes uses JPEG 2000 inside, or a print-ready PDF may store photos as CMYK. Both are deliberately skipped because the tool does not know if changing them would alter the rendered colors. The honest result is "no change" rather than a risky rewrite. If you have ever wondered why some PDFs refuse to shrink no matter what you do, it is usually because the images inside are not the JPGs this tool can touch.
Limits you should know before you start
The tool is bounded on purpose. The current build accepts one PDF up to 25 MB, processes up to 80 eligible image objects, and caps the total decoded pixels at 100 megapixels across the images it reads. A single 4000 × 3000 JPG is 12 megapixels, so eight such images sit at 96 megapixels, comfortably inside the cap. A ninth would push past 108 megapixels and the tool will stop. These caps exist so the work stays responsive on a normal laptop and so the browser does not run out of memory mid-rewrite.
| Limit | Value |
|---|---|
| Maximum input file size | 25 MB |
| Maximum eligible image objects | 80 |
| Maximum total decoded pixels across eligible images | 100 megapixels |
| Maximum longest side after resize | 1400 px |
| JPEG export quality | 0.70 |
| Eligible image color spaces | DeviceRGB, DeviceGray (unmasked DCTDecode only) |
Tips for getting the smallest reliable result
Because the tool only acts on eligible JPGs, the size of your savings depends on what is in the file before you start. A few habits help you get the most out of it without surprises:
- Keep the original PDF until you have inspected the output. The tool is non-destructive to the source, but you still want a clean copy to fall back on in case the visual result is not what you expected.
- Open the output before sharing it. The tool verifies page count and that the PDF opens cleanly, but a quick visual check on the heaviest pages is still worth a minute. Look at the biggest photo on each spread and confirm it is still sharp enough for its purpose.
- Use a desktop PDF optimizer for archival, print, or color-managed work. This browser tool is intentionally narrow; it is not designed for guaranteed lossless compression, archival conformance such as PDF/A, or strict print color management. For those jobs, a specialist tool is the right answer.
- Pair it with the right companion tools. If the file is still too big after one pass, try Split PDF to break it into chunks an email system will accept, or Resize PDF to drop page dimensions before you compress. If the file is mostly scanned pages that should look like clean images, PDF to JPG Converter or PDF To PNG can give you a smaller image-based version of the content.
The point is that Compress PDF stays small and honest about its limits. If your PDF has the kind of JPG-heavy content it is built for, you get a smaller, visually faithful file in one step, with nothing leaving your browser.